By Deborah Borfitz

March 22, 2021 | AstraZeneca has been embracing the use of real-world evidence (RWE) for clinical research to help overcome some of the new and perennial challenges in moving medicines to market, according to Xia Wang, Ph.D., director of health informatics, during a presentation made at the recent Summit for Clinical Ops Executives (SCOPE). The company has been actively using RWE to optimize study protocols, assess new opportunities, employ external arms/comparators, enable pragmatic clinical trials, and transform endpoints, she reports.

The landscape for clinical trials is beset with challenges, ranging from complex study design and the search for precision patient populations to fierce competition with outside companies and even within the same organization when a product has multiple applications, says Wang. Trials are increasingly costly to conduct, longer in duration, and at heightened risk of termination, while investigators and sites are seeking to satiate their appetite for more innovative trials.

The global pandemic, meanwhile, has been a “double-edged sword” adding complexity but also forcing speed and the rapid adoption of digital technologies, she adds.

Patients are the central focus of value-based healthcare and, as such, at the center of everything done by pharmaceutical companies, agency policies, and healthcare payers and providers, Wang says. When all parties are working together, the benefits of patient-centered medicine development extend across the entire ecosystem.

To Wang, real-world data (RWD) is all data generated outside of a randomized controlled trial (RCT) and includes electronic health records (EHRs), patient surveys, and social media data. RWE is the clinical evidence about the usage and real-world potential benefits and risks of a medical product derived from RWD.

The U.S. Food and Drug Administration (FDA) has made major efforts to steer discussions around RWD and RWE for clinical research purposes. Many workshops have been held, and regulatory guidelines issued, over the past few years, she notes. In December 2018, the agency published its RWE program framework.

Among the pragmatic trials that have been discussed at FDA public meetings is GlaxoSmithKline’s phase 4 Trelegy effectiveness trial for chronic obstructive pulmonary disease (COPD) that completed five years ago. It was a “beautifully” designed study but complicated in design and its data collection, Wang says, in addition to be costly to conduct.

Pragmatic Trials

AstraZeneca has been leveraging the ready availability of federated EHRs (e.g., TriNetX, InSite, and Clinerion), providing “easy access to almost real-time” aggregated patient data to evaluate study eligibility criteria, says Wang. For one COPD trial, for example, AstraZeneca used federated EHRs to assess how different backup treatments impacted trial design and the segmenting of populations into the study.

For protocol optimization purposes, AstraZeneca has also used Optum claims data to see the effect on patient enrollment in a continuous triple-therapy treatment for COPD— inhaled corticosteroid, a long-acting β2-agonist, and a long-acting muscarinic antagonist—over a one-year analysis timeframe, says Wang.

Additionally, AstraZeneca has used MarketScan data on severe uncontrolled asthma to develop an external (comparable) study arm reflective of the inclusion and exclusion criteria in a phase 3 study to “contextualize safety events,” Wang says. Three databases—MarketScan, IQVIA, and Aetion—provided a sample size large enough to learn what kinds of events were even possible to contextualize, she adds. 

Multiple factors need to be considered for pragmatic clinical trials blending traditional RCTs with RWE, says Wang. The framework adopted by AstraZeneca includes defining the patient population and expected uses cases, and identifying which database meets the needs of the required patient profile and then querying that database for eligible patients.

Study teams also need to think about data network providers and how they might be approached about collaborating, she continues. Data collection and visits need to parallel real-world practice but “keep the rigor” of a RCT so safety signals are detected.

Traditional, pragmatic, and innovative study design elements have been incorporated into the Dapagliflozin Effects on Cardiovascular Events in Patients With an Acute Heart Attack (DAPA-MI) trial, which AstraZeneca launched late last year in Sweden and the U.K. It was the world’s first indication-seeking, registry-based randomized controlled outcomes trial, and was underpinned by digital technologies.

The study is tapping two registries in the U.K. (MINAP) and Sweden (SWEDEHEART), and many data elements are flowing from the registries to the study database right away, she says. The hope is that this will reduce study-related costs.

Lessons Learned

It is important that research endpoints to be collected in real-world practice translate easily into those environments, she says. Labs and hospitalizations are easy enough to capture in the clinic, but some composite scores (e.g., SRI4 in lupus trials) are not. In these instances, machine learning and artificial intelligence might be used to identify predictors for treatment responders that can be easily translated or validated with RWD for clinical decision support. In the future, patient-reported outcomes and data collected from wearable devices could also be developed into endpoints.

Among Wang’s observations are the rapid adoption of RWE in clinical study design and protocol optimization and lack of seamless connections to implement site identification and trial recruiting activities. She notes a “strong scientific desire to link real-world evidence with insights from other data sources [e.g., imaging and genomics data] to empower or improve the confidence in the new target identification.”

Wang says the “appetite for innovation” is strong across the spectrum of real-world trials. But given the complexity of their design and execution, internal and regulatory buy-in are essential with the understanding that “before you spend less you may have to spend more.”

Interest is also high in piloting machine learning and artificial intelligence applications with RWE to identify diseases early and prevent their progression, says Wang. “At the moment, we need more clarity on what an end-to-end machine learning/AI product looks like and how this product can be sustainably maintained over time when new data feed in so you can continue to improve the model and get a better prediction.”